Cargando…

A Variational Stacked Autoencoder with Harmony Search Optimizer for Valve Train Fault Diagnosis of Diesel Engine †

Diesel engine fault diagnosis is vital due to enhanced reliability and economic efficiency requirements. The extracted features in traditional fault diagnosis are constructed manually, which is very cumbersome because of the requirement for lots of expertise. To handle this issue, this paper propose...

Descripción completa

Detalles Bibliográficos
Autores principales: Chen, Kun, Mao, Zhiwei, Zhao, Haipeng, Jiang, Zhinong, Zhang, Jinjie
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2019
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6982694/
https://www.ncbi.nlm.nih.gov/pubmed/31906062
http://dx.doi.org/10.3390/s20010223
Descripción
Sumario:Diesel engine fault diagnosis is vital due to enhanced reliability and economic efficiency requirements. The extracted features in traditional fault diagnosis are constructed manually, which is very cumbersome because of the requirement for lots of expertise. To handle this issue, this paper proposed a variational stacked autoencoder (VSAE) to adaptively extract features from angular domain signals. As an unsupervised algorithm, VSAE can extract high-level features with the help of multiple encoding layers. Layer-wise pre-training and fine-tuning are introduced to get a better network initialization value. Moreover, the dropout technique and the batch normalization technique are carried out to prevent over-fitting and implement fast convergence. Finally, the harmony search optimizer (HSO) algorithm is introduced to get an appropriate hyper-parameter setting in the VSAE model, as well as make adaptive adjustment of the network structure. In order to verify the proposed method, the valve train fault data is collected on the diesel engine test rig under twelve operating conditions. The results indicate that the proposed scheme can effectively diagnose different degrees of intake valve fault, exhaust valve fault, and coupling fault under various operating conditions. Furthermore, the classification accuracy improved from 94.10% to 98.85%VSAE compared with stacked autoencoder (SAE) and some other traditional fault diagnosis algorithms.